Technical SEO · guide 7 of 7
Core Web Vitals explained without the jargon
What LCP, INP and CLS really measure, Google’s published “good” lines, why lab scores and real visitors disagree, and how much speed matters for ranking.
25 min readUpdated 9 Nov 2026
- 9 questions to test yourself
- 1 calculator
- 11-item checklist
On this page13 sections
A perfect score and a failing grade on the same page
Your developer runs a speed test and shows you a green 100 out of 100. A week later, Google Search Console emails you: your pages have “poor” Core Web Vitals on mobile. Somebody is wrong. Or are they?
Both can be true at once, and once you see why, most of the confusion around site speed disappears. By the end you will know what the three numbers measure, which of your test results Google actually uses, how much speed really affects ranking (less than many people selling speed work suggest), and exactly which fixes to ask a developer for.
Two things along the way surprise almost every owner. Your average visitor does not decide your grade, and one of the cheapest fixes on the web has nothing to do with making anything faster. We will get to both. First, your instinct.
Quick quiz
A page scores 100 in a speed test, but Search Console says it fails Core Web Vitals on mobile. What is the most likely explanation?
Pick one. Your first answer is the one that counts.
Three questions every visitor silently asks
Forget the acronyms for a moment. Think of walking into a restaurant. You want three things: the menu arrives quickly, the waiter reacts when you raise your hand, and the table does not slide away while you are reaching for your glass. Core Web Vitals are those three things for a web page.
Largest Contentful Paint (LCP) is the menu arriving. It measures how long it takes for the biggest thing in the first screen, usually the main photo or the headline, to appear. It answers “is anything happening?”
Interaction to Next Paint (INP) is the waiter. It measures how long the page takes to visibly react after you tap a button, open the menu or type in a form. It watches the interactions during the whole visit and reports one of the slowest. It answers “is this thing listening to me?” INP replaced an older metric, First Input Delay, as a Core Web Vital in March 2024, so older articles that mention FID are out of date.
Cumulative Layout Shift (CLS) is the sliding table. It scores how much things jump around while you are looking at them: the text you were reading drops down because a banner loaded above it, or the button you were about to tap moves and you hit an advert instead. It answers “can I trust where things are?”
Now watch all three go wrong in six seconds. Step through one bad page load below: the page is made up, but every moment in it happens on real small-business sites.
Step by step
An illustrative example: a phone loads a made-up plumber’s homepage
Step 1 of 6
0.0 seconds: the tap
Someone taps your result in Google on a mid-range phone with a patchy mobile signal. The clock for LCP starts now, not when your server answers.
Google’s “good” lines for each metric
Google publishes exact thresholds for each metric, so you never have to guess what “good” means. Before you read them, try one.
Make a guess
How quickly does the main content need to appear for Google to call LCP “good”?
Anything within ±0.5 seconds counts as right.
Side by side
Google’s published thresholds
| Metric | Good | Needs improvement | Poor |
|---|---|---|---|
| LCP (loading) | 2.5 s or less | 2.5–4 s | Over 4 s |
| INP (reacting to taps) | 200 ms or less | 200–500 ms | Over 500 ms |
| CLS (stability) | 0.1 or less | 0.1–0.25 | Over 0.25 |
The 75% rule: your slowest customers set the grade
Here is the first of the two surprises. Before reading the rule, try to apply it.
Quick quiz
An illustrative example: in the last 28 days a page had 100 phone visits. 70 saw the main content in 1.5 seconds and 30 waited 3.2 seconds. The average is about 2 seconds. Does LCP pass?
Pick one. Your first answer is the one that counts.
That is the 75% rule, and it changes how you read every report. Google lines up the real visits from the last 28 days and reads the value at the 75th percentile, separately for phones and for desktops. A metric is “good” only if that value is inside the good line.
Put plainly: your slowest quarter of customers decide the grade. For a local business whose customers are on the bus, at a building site or in a village with one bar of signal, that quarter is not a rounding error. It is also why testing on your office Wi-Fi tells you so little.
Lab data versus field data: why the two disagree
Every speed report you will ever see is one of two kinds, and mixing them up causes most of the arguments, including the one from the first screen of this guide.
Lab data is a rehearsal. A tool such as Lighthouse, which powers the bottom half of PageSpeed Insights, loads your page once on a simulated mid-range phone with a deliberately slowed connection, and reports what it saw. It is repeatable and instant, which makes it ideal for a developer checking whether a change helped.
Field data is the real performance. It comes from the Chrome User Experience Report (CrUX): anonymous measurements from real Chrome users who have opted in to sharing usage statistics, collected over a rolling 28 days. This is the data behind Search Console’s Core Web Vitals report and the top half of PageSpeed Insights, and it is the data Google’s ranking systems use.
Side by side
| Question | Lab data (Lighthouse test) | Field data (real visitors, CrUX) |
|---|---|---|
| Who is measured | One simulated visit | Real Chrome users who opted in |
| When it updates | Every time you run the test | Rolling 28 days, so fixes show gradually |
| Measures INP | No: nobody taps. Total Blocking Time stands in | Yes, from real taps, clicks and key presses |
| Sees shifts after scrolling | Only during the initial load | Yes, for the whole visit |
| Gives the 0–100 score | Yes | No: good, needs improvement or poor |
| Used by Google for ranking | No | Yes, as part of page experience |
| Best for | Finding causes and checking a fix | Knowing what customers live through |
Pass or fail? Grade five reports the way Google does
In PageSpeed Insights, the field data at the top ends in a single line: “Core Web Vitals Assessment: Passed” or “Failed”. A page passes when all three metrics are good at the 75th percentile. If there is not enough INP data, it is judged on LCP and CLS. You now know enough to grade pages yourself: try these made-up results.
Myth or fact?
True or false? Grade each page the way Google does
1.LCP 2.3 s, INP 180 ms, CLS 0.08: this page passes.
2.LCP 1.2 s, INP 350 ms, CLS 0.01: this page passes, because two of three are excellent.
3.LCP 2.6 s, INP 90 ms, CLS 0.02: this page fails.
4.Lighthouse performance score 98, field data CLS 0.31: this page passes, because the score is green.
5.A page with too few visitors for field data has neither passed nor failed.
How much Core Web Vitals matter for ranking, honestly
Before the honest answer, a decision that real owners get offered all the time. What would you do?
What would you do?
An illustrative example: the €1,800 speed offer
A made-up florist’s site shows “Passed” in the field data on mobile and desktop. The lab score is 82. An agency offers a €1,800 “speed optimisation package” to reach 100, “so Google ranks you higher”. Meanwhile the site has no page for wedding flowers, the florist’s most profitable service.
What would you do with the budget?
Google does use Core Web Vitals in ranking. Its Search Central documentation on page experience says so, and recommends site owners achieve good scores. The same documentation is unusually blunt about the limits: Google’s systems try to show the most relevant content even if its page experience is sub-par, and good Core Web Vitals do not guarantee a top position.
In practice, speed works like a tiebreaker. Google has described it in similar terms: when many pages are about equally relevant to a search, page experience can make more of a difference. A slow page with the best answer, the most reviews and the strongest reputation can still outrank a fast page that says little. Moving from “needs improvement” to “good” is worth doing; chasing a lab score from 90 to 100 is rarely worth paying for. The guide on how Google decides who ranks first, in plain English, shows where speed sits among the bigger signals.
The bigger payoff is usually not Google at all. Slow, jumpy pages lose people who have already found you: the visitor who gives up before the photo loads, or taps the wrong button and leaves. Google’s own Think with Google research has long made the point that mobile visitors abandon slow pages. That cost is real whether or not your ranking moves.
If a package is sold with the words “so Google ranks you higher”, ask the question from the scenario above: which field metric fails? The guide on SEO scams and red flags lists the other promises worth that same question.
The usual culprits on small-business sites
Most small-business sites that fail do so for the same handful of reasons. None is exotic, and you can often spot them yourself by loading your own homepage on your phone with Wi-Fi switched off. Before each card flips, guess which of the three metrics it hurts: LCP (slow to appear), INP (slow to react) or CLS (things jump).
Test yourself
Which metric does it hurt? Guess, then flip
Think of your answer, then flip the card.0 of 8 flipped
Two of the eight cards are about photos, and on many small-business sites the hero photo alone decides LCP. The guide on image SEO shows how to find your heaviest image in two minutes and what size and format to ask for.
The cheapest fix on the web: stop the page jumping
Here is the second surprise promised at the start. Layout shift is easiest to understand when you imagine it happening to you, so picture the same moment on two versions of a made-up booking page.
Before and after
An illustrative example: tapping “Book a table” on a phone
Space not reserved
You read the opening hours and move your thumb to “Book a table”. Just as you tap, a photo with no set size finishes loading above it and pushes everything down by a third of the screen.
You tap “Gift vouchers” instead. You go back, find your place again, and wonder if the site is broken.
Space reserved
The photo’s box is drawn at its final size from the start, with a soft grey background. When the photo arrives, it fills the box. Nothing moves.
You tap “Book a table” and land on the booking form, first time.
- The only change was telling the browser each image’s width and height in advance, so it can save the space before the photo arrives. It costs a developer minutes, and it is often enough on its own to bring a failing CLS back under 0.1.
- Both versions load the photo at exactly the same speed. CLS is about trust, not speed: whether things stay where you saw them.
Fixes you can ask a developer for
You do not need to make these changes yourself, but naming them precisely gets you a quote for specific work instead of a vague “speed package”. Match each request to the metric that fails in your field data, and ask what the developer will measure before and after.
To see these requests applied in order, read the Fenerche Lamps study on our Case studies page. It is an illustrative example: a made-up lamp shop whose camera-sized product photos, render-blocking scripts and missing caching left mobile LCP at 7 seconds in the field data, and the fixes that brought it under the 2.5-second line one at a time.
Checklist
Requests worth making, grouped by metric
Saved on this device only
Before you test: predict your own result
Answer these from what you already know about your site. In the next section you will run the real test, so this is a chance to see whether your gut feeling matches what your customers actually experience.
Score yourself
Your site’s speed risk, point by point
The homepage has a slider, background video or a very large photo at the top
A chat, booking or review widget loads on every page
A cookie banner or promo bar pushes the page down when it appears
Photos are uploaded straight from a camera or phone without resizing
The site was built with a heavy theme or page builder and several plugins
Nobody has looked at the field data in PageSpeed Insights or Search Console this year
0 of 6 answered. Answer them all to see your result.
Check your own pages in ten minutes
Now check your guess. You can see your real grade today, for free, without logging in to anything technical. The free speed test at the end of this guide runs the same Google measurement and explains each number in plain words; PageSpeed Insights is Google’s own version of the same report.
- Test your homepage and your most important service or product page, not just the homepage, in the speed test below or in PageSpeed Insights (pagespeed.web.dev).
- Read the top section first: “Discover what your real users are experiencing”. That is the field data. Note the assessment and the three numbers for mobile, then switch to desktop.
- If it says there is not enough data for the page, look at the result for the whole site (“Origin”), if shown.
- Only then scroll to the lab diagnostics, which list likely causes such as “Largest Contentful Paint element” and “Avoid large layout shifts”. Screenshot them for your developer.
- If you have Search Console, open Core Web Vitals in the menu. It groups similar pages, so one fix to a template can clear dozens of URLs at once. The Search Console guide for business owners shows where this report sits among the others worth a monthly look.
- After a fix ships, re-run the lab test straight away to confirm it, then give the field data about four weeks to catch up.
Try the numbers
Type in your field results: how far are you from “good”?
Use the mobile field numbers from the top of PageSpeed Insights or from Search Console. If the first result is 0, your speed work is done for now. Otherwise, go back to the developer checklist above and tick only the lines for the metric that is short.
Metrics outside the “good” line
2 of 3 to fix
LCP: time to shave off to reach 2.5 s
0.7 s
INP: delay to cut to reach 200 ms
50 ms
This reads your own numbers against Google’s published lines; it is not a forecast of rankings or results. CLS has no gap shown here: above 0.1 needs work, above 0.25 is poor, and the CLS lines in the checklist are usually the quickest wins.
Questions owners ask, and where to go next
Questions people ask
I fixed the problem. Why does Search Console still say “poor”?
Field data covers the last 28 days of visits, so old slow visits keep counting until they roll out of the window. In Search Console you can click “Validate fix” to have Google track the change. Expect around four weeks.
Is my Lighthouse score used for ranking?
No. Google’s ranking systems use field data from real users. The lab score is a diagnostic tool for finding causes and checking fixes.
My page has no field data. Does Google ignore its speed?
When there is not enough data for a single page, Google can still have data for similar pages or for the whole site. Either way, a fast page is better for the visitors you do have, so the lab test is still worth using as a guide.
Should I switch platform to get better Core Web Vitals?
Rarely as a first step. Most failures come from images, widgets and banners, which exist on every platform. Fix those first; consider a rebuild only if the theme itself is the bottleneck. If you do, the guide on choosing WordPress, Shopify, Wix or custom compares the options, and the guide on moving or redesigning a site without losing your Google traffic shows how to keep your URLs.
Desktop passes but mobile fails. Which one counts?
Both are reported separately, and for most local businesses most visitors arrive on phones (your analytics will tell you your own split). Treat mobile as the one to fix first. Mobile failing while desktop passes is normal: phones have slower processors and connections.
Your result from the calculator points to your next read.
- LCP over 2.5 s: read the guide on image SEO next. The hero photo is the usual cause, and it is the fix you can check yourself the same day.
- All three good: your speed work is done for now. The guide on writing a page that deserves to rank is where the next budget usually pays back more.
- No field data yet: the guide on Google Search Console for business owners shows how to watch the report fill in as visits grow.
- A redesign or new theme coming: the guide on planning a website redesign helps you keep what works, so the new site does not start with a slider and five widgets.
You reached the end
Your summary
This guide has 9 quick questions
Try them to see how much you already know. Your first answer is the one that counts.
Technical SEO: 0 of 7 read
- 1. Why isn't my page on Google? 12 causes and fixes: not read yet
- 2. Schema markup: what it does and which types a small business needs: not read yet
- 3. Multilingual websites: reach customers in every language without confusing Google: not read yet
- 4. Duplicate content and canonicals: what Google really does with copies: not read yet
- 5. Image SEO: faster photos that Google can find and understand: not read yet
- 6. Moving or redesigning a site without losing your Google traffic: not read yet
- 7. Core Web Vitals explained without the jargon: this guide, not read yet
0 of 57 guides read · Saved on this device only
New guides by email when they are published.
Unsubscribe in one click.
What to do next
- Run PageSpeed Insights on your homepage and your top service page, and write down the mobile field LCP, INP and CLS.
- For any metric outside the good line, pick the matching requests from the checklist and send them to your developer with a screenshot.
- Load your homepage on your own phone with Wi-Fi off and watch for a late photo, a jump or a tap that does nothing.
- If all three metrics are already good, stop paying for speed work and put that budget into pages that answer your customers’ searches.
How fast is your site, really?
Test your pages on a simulated phone and see the Core Web Vitals Google measures.
Run the speed testNext in Technical SEO
Technical SEO
Image SEO: faster photos that Google can find and understand
The photo you are proudest of may be the reason your homepage feels slow. Here is how to check in two minutes.
26 min readStart reading
See it applied in an illustrative case study
Invented businesses and modelled figures, not client results: how the work in this guide fits into a whole engagement.
- Signal · 10 monthsIllustrative+168% holiday-club bookings from searchHazelnook Forest School
How a Brecon forest school pruned 3,000 booking-date pages and got its own programmes found
Forest school & holiday club, 12 leaders · Brecon · UK & Ireland
- Signal · 6 monthsIllustrativeMobile conversion rate +64%Hrimfaxi Tours
How a Reykjavík tour operator fixed a slow site that was losing aurora bookings
Northern lights & glacier tours · Reykjavík · Europe